Quantcast Detection of erroneous data

Order this information in Print

Order this information on CD-ROM

Download in PDF Format


Click here to make tpub.com your Home Page

Page Title: Detection of erroneous data
Back | Up | Next

tpub.com Updates




Information Categories
.... Administration
Food and Cooking
Nuclear Fundamentals
  Educational CD-ROM's
Printed Manuals
Downloadable Books


Back ] Home ] Up ] Next ]

Click here to Order your Radar Equipment Online

DETECTION OF ERRONEOUS DATA.— The validity of a report or part of a report becomes suspect when it (1) is inconsistent with nearby reports (in case of dense synoptic net-works), (2) contains internal inconsistencies, or (3) leads to marked or unlikely changes in con-tinuity or history (in areas of sparse data). With the first two cases, the report should be compared with neighboring reports exercising care that the elements you’re comparing were ob-served at the same altitude and over the same type of underlying surface. The last case involves an isolated report and requires the use of all your analytical tools. If possible, the previous 3-hourly reports from the station in question should be con-sulted to check for continuity. Sometimes, in the case of an apparently erroneous isolated report, it is virtually impossible to check its validity. However, never disregard or discard it simply because it is difficult to fit into a preconceived pattern. At least one or more valid logical reasons should exist for not drawing to such a report. Many analysts have violated this rule only to find that subsequent charts confirmed the existence of an important meteorological event.  

Climatology and common sense should also be used to detect erroneous reports; for example, you wouldn’t expect an 80°F temperature inside the Arctic Circle. Random errors are the most difficult to detect unless they are large, while systematic or repeated errors are more easily discovered and corrected. Also, as the number of errors or inconsistencies in a single report increases, the entire report must become suspect. 

CLASSIFICATION AND CORRECTION OF ERRORS.— Once an error is detected, the data must not be discarded arbitrarily. Some attempt should be made to estimate the proper correction of the data. Consider the possible sources of the error. The sources are classified as follows: errors due to encoding, transmission, decoding, and plotting; observation errors; and computational errors in data not directly observed.

Errors in the first group are essentially com-munications errors. Frequently, on teletypes an incorrect number is substituted for the correct one; for example, a five or a seven for a six. This type error is equally probable in any numerical group but will go undetected unless overly obvious. Thus, errors in the tens digit of pressure or temperature are usually spotted while similar errors in the units digit cannot be identified so easily. Errors of this type are also common in wind direction and ship position reports. A ship’s recorded position is oftentimes questionable, because it was encoded and/or transmitted incor-rectly, or it was misplotted. In either case, you’ll normally find the position off by 5 to 10 degrees of latitude or longitude. A good remedy for posi-tion and other type errors from ships is to keep a running log of ship reports. Plotting the ship’s reports side by side permits easy comparison of suspect data. Mistakes in decoding or plotting are the most easily checked, requiring only a check of the original message. This should always be the first step in questioning inconsistent or suspect data. Plotting of off-time reports on the current chart is another common error. For example, a few 12Z reports may show up within a block of 18Z reports and be inadvertently plotted. Detect-ing this error may or may not be easy.

Reports containing observational and com-putational errors are almost impossible to correct. However, when stations make consistent, obvious errors in weather element values, maintain a list of these stations, the elements in error, and the probable applicable corrections.  

Computational errors are found in sea level pressure reductions, dew-point temperatures, and true wind speed and direction reported by moving ships. The first two elements are generally taken directly from tables or graphs. Errors arise principally from misreading the tables or graphs. In the case of true wind, greater reliance should be placed on reports from ships known to carry well-trained observers, such as weather ships, naval vessels, and large commercial liners. Many smaller vessels are not equipped with anemom-eters, in which case direct estimates of true wind are made using wind wave vs wind speed tables.

Back ] Home ] Up ] Next ]


Privacy Statement - Press Release - Copyright Information. - Contact Us - Support Integrated Publishing

Integrated Publishing, Inc.